Download Data Driven Data Mining to Domain Driven Data

Global Journal of Computer Science and Technology
Volume 11 Issue 23 Version 1.0 December 2011
Type: Double Blind Peer Reviewed International Research Journal
Publisher: Global Journals Inc. (USA)
Online ISSN: 0975-4172 & Print ISSN: 0975-4350
Data Driven Data Mining to Domain Driven Data Mining
By Mitu Kumari
Kurukshetra University, Kurukshetra, Haryana, India
Abstract - In the preceding decade data mining has came into sight as one of the largely
energetic areas in information technology. Traditional data mining is seriously dependent on data
itself, and relies on data oriented methodologies. So, there is a universal necessity in bridging the
space among academia and trade is to provide all-purpose domain-related matters in
surrounding real-life applications. Domain-Driven Data Mining try to build up general principles,
methodologies, and techniques for modelling and reconciling wide-ranging domain-related
factors and synthesized ubiquitous intelligence adjacent problem domains with the data mining
course of action, and discovering knowledge to hold up business decision-making.
Keyterms : Data Mining, Domain driven data mining, decision-making.
GJCST Classification : H.2.8
Data Driven Data Mining to Domain Driven Data Mining
Strictly as per the compliance and regulations of:
© 2011I . Mitu Kumari. This is a research/review paper, distributed under the terms of the Creative Commons AttributionNoncommercial 3.0 Unported License http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use,
distribution, and reproduction inany medium, provided the original work is properly cited.
Data Driven Data Mining to Domain Driven Data
Mining
Keywords : Data Mining, Domain driven data mining,
decision-making.
I
I.
INTRODUCTION
n the last ten years, data mining is a field which
becomes the most active, dynamic and lively area in
information and communication technologies. The
rapid growth of the global economy and heavy usage of
computing and networking across every sector and
business, results in data and its deep analysis becomes
a particularly important issue for the soft control of an
organization , and also equally important for the
production system , decision making powers and
performance of the organization. Now these days, there
is a rapid increase in the applications of the data mining
in various fields like business, government, social
networks and the like ones. But due to the data driven
data mining’s limited decision support power in the real
world, it hinders from playing a strategic decision
support role in all these areas. In order to sort out this
problem, a new approach Domain Driven Data Mining is
evolved, this new approach will handle all the issues
which are faced by the traditional data mining and also
tackle the findings, thoughts and lessons learned in
conducting several large scaled real world data mining
business applications. The motivation of Domain Driven
Data Mining is to study effective and efficient
methodologies, techniques, tools and applications that
can discover and deliver actionable knowledge that can
be passed directly to the business people for the direct
decision making and action taking.
If we apply current data mining algorithms and
techniques on to the real world problem solving and
Author : Kurukshetra University, Kurukshetra, Haryana, India.
E-mail : [email protected]
decision making tasks then we have to face the crucial
need to lessen the differences between academic world
and commerce. Also we have to tackle the space
between estimation systems and real business
requirements. Not only this, we also have to manage the
inequilibrium between the huge number data mining
algorithms existing in the market in opposition to those
few data mining algorithm that are in fact deployed in
problem areas and resulting in those patterns which are
of real use and these patterns can be suggested for
decision support actions.
Real world data mining applications have
projected critical desires for discovering actionable
knowledge of foremost interest to real user and
business wishes. As the actionable knowledge
discovery is noteworthy and also very demanding.
In order to overpass the above mentioned gaps,
it is vital to boost the decision support power of data
mining and knowledge findings. It is crucial to expand
the actionability of the discovered patterns and to make
available the results that can sustain decision making, in
the right and beneficial route.
Domain driven data mining provides an efficient
overview of the issues in discovering actionable
knowledge and advocates the methodology of mining
the actionable knowledge in constraint based context
through human mining cooperation in a loop closed
iterative improvement manner. It is valuable for
promoting the paradigm shift from data driven hidden
pattern mining to area driven actionable data discovery.
Further, progress in studying domain driven data mining
methodologies and applications can facilitate the
deployment swing from standard and artificial data set
based testing to genuine data and business
atmosphere based back testing and development.
II.
DATA DRIVEN DATA MINING
A distinctive feature of traditional data mining is
that KDD (Knowledge Discovery From Data). One of the
elementary objectives of KDD is to discover knowledge
that is of key concentration to genuine business
requirements and user preferences but KDD is a
presumed and preset process. It targets the production
of predefined and automatic algorithms and tools. As a
consequence, the algorithms and tools developed have
no potential to adapt to external environment
constraints. Millions of patterns and algorithms have
© 2011 Global Journals Inc. (US)
65
Global Journal of Computer Science and Technology Volume XI Issue XXIII Version I
Abstract - In the preceding decade data mining has came into
sight as one of the largely energetic areas in information
technology. Traditional data mining is seriously dependent on
data itself, and relies on data oriented methodologies. So,
there is a universal necessity in bridging the space among
academia and trade is to provide all-purpose domain-related
matters in surrounding real-life applications. Domain-Driven
Data Mining try to build up general principles, methodologies,
and techniques for modelling and reconciling wide-ranging
domain-related factors and synthesized ubiquitous intelligence
adjacent problem domains with the data mining course of
action, and discovering knowledge to hold up business
decision-making.
December 2011
Mitu Kumari
Data Driven Data Mining to Domain Driven Data Mining
Global Journal of Computer Science and Technology Volume XI Issue XXIII Version I
December 2011
been available in literature but sorry to say that a small
number of them have been transferred into real
business.
In data driven data mining lots of patterns are
generated according to the problems but they are not
enlightening and clear to business individuals. They
can’t straightforwardly acquire truly remarkable and
operable patterns for their business. A large fraction of
the indentified patterns may be either commonsense or
of no particular attention to business desires. Business
grassroots are puzzled as to why and how they should
be concerned regarding those conclusions. Activities
extracted or summarized through post investigation and
post processing without in view of business concerns do
not replicate the authentic expectations of business
desires. Therefore they cannot bear smart decision
making. Business people often don’t know and also not
66 well-versed regarding, how to understand and utilize the
discovered patterns and what undemanding activities
can be taken to engage those discovered patterns in
business functioning systems and decision making.
Conventional KDD is a data centred and
technically dominated course targeting automated
hidden pattern mining. The core objective of
conventional data mining research is to let data verify
research innovation, track the elevated performance of
the algorithms and express novel algorithms. As a
consequence, the mining process stops at discovering
knowledge that is primarily of importance to academic
or industrial individuals.
In the real world, determining and transporting
knowledge that is actionable in answering business
problems has been analyzed as the fundamental nature
of KDD. However, the existing data mining is principally
data-centred and technically conquered, and stops at
hidden pattern mining favouring technical concerns and
expectation, while many other features surrounding
business problems have not been thoroughly or
exhaustively considered and balanced. It will be one of
the great challenges to the existing and future KDD
society.
A distinctive fashion in real world data mining
applications is to treat a data mining system as a
problem solving systems within a certain atmosphere.
Looking at the problem solving from the domain driven
point of view, a lot of unwrap matters and opportunities
arise, demonstrating the need of next generation data
mining and knowledge discovery far further than the
data mining algorithms themselves. In order to sort out
these troubles a new methodology is proposed i.e.
Domain driven data mining. Domain driven data mining
tends to create next generation methodologies,
techniques and tools for a probable idea shifting from
data driven hidden pattern mining to domain driven
actionable data delivery.
© 2011 Global Journals Inc. (US)
III. DOMAIN DRIVEN DATA MINING
Intending at complementing the inadequacy of
conventional data mining, in particular, reinforcing the
problem-solving-oriented ability and deliverables in
enterprise data mining, we recommend a realistic
methodology, called Domain Driven Data Mining by
following the extensively acknowledged jargon ‘Data
Mining’.
Domain Driven Data Mining is proposed as a
methodology and a collection of techniques targeting
domain driven actionable knowledge delivery to drive
Knowledge Discovery from Data (i.e. KDD) toward
enhanced
problem-solving
infrastructure
and
capabilities in real business state of affairs. On top of
the data-centred framework, Domain Driven Data Mining
aims to build up proper methodologies and techniques
for targeting domain knowledge, human role and
interaction, organizational and social factors, as well as
capabilities and deliverables toward delivering
actionable knowledge and supporting business
decision-making action taking in the KDD process. In
other words we can say that Domain Driven Data Mining
intend to create subsequent generation methodologies,
techniques and tools for a probable paradigm transfer
from data centred out of sight pattern mining to domain
driven actionable knowledge delivery.
As a result of the Domain Driven Data Mining
investigation and development, we can deliver businessfriendly and decision-making rules and actions that are
of solid technical and business importance.
“Domain driven data mining refers to the set of
methodologies, frameworks, approaches, techniques,
tools and systems that deliver for human, domain,
organizational and social, and network and web factors
in the environment, for the innovation and delivery of
actionable knowledge. Actionable knowledge means
business responsive and comprehensible, reflects user
preferences and business needs, and can be
effortlessly taken over by business individuals for
decision-making and action-taking”
The existing data mining methodology, usually
chains self-governing pattern discovery from data. By
contrast, the suggestion of domain driven knowledge
discovery is to engage ubiquitous intelligence into data
mining. The Domain Driven Data Mining highlights a
procedure that discovers in-depth patterns from a
constraint-based environment with the contribution of
domain specialists and their acquaintance. Its intention
is to maximally accommodate equally naive users as
well as practised analysts, in addition to satisfy business
goals. The patterns discovered are expected to be
integrated into business systems and to be aligned with
existing business rules. To formulate domain driven data
mining successful, user guides and intellectual humanmachine interaction interfaces are indispensable
through incorporating mutually human qualitative
aptitude and machine quantitative aptitude. In totalling,
appropriate mechanisms are obligatory for dealing
IV. KEY ELEMENTS OF DOMAIN DRIVEN
DATA MINING
In domain-driven data mining, the following
seven key elements play a very important role. They
have capability of building a KDD which is dissimilar
from the alive data-driven data mining if they are
properly considered and supported from technical,
procedural, and business point of view.
a) Restraint -Based framework
In human society, everyone is restrained either
by communal regulations or by individual situations.
Similarly, actionable knowledge only can be discovered
in a restraint-based framework such as environmental
authenticity, opportunities, and restraints in the mining
procedure. Particularly, in the first section, we catalogue
some types of restraints that play noteworthy roles in a
process, effectively discovering knowledge actionable to
business. In practice, a lot of other aspects, such as
data stream and the scalability and effectiveness of
algorithms, may be enumerated. They consist of
domain-specific, functional, nonfunctional, in addition to
environmental restraints. These ubiquitous restraints
create a restraint-based framework for actionable
knowledge discovery. All of the preceding restraints to
varying degrees have to be considered in significant
phases of real-world data mining. In this case, it is even
called restraint-based data mining.
b) Incorporate Field Awareness
It is accepted gradually that field awareness can
play noteworthy roles in real-world data mining. For
instance, in trade (buy and sell) pattern mining, brokers
often take “beating market” as an individual liking to
judge a recognized rule’s actionability. In this case,
stock mining system requires to set in the formulas
calculating market return and rule return, and set an
interface in order for traders to specify a most wanted
threshold and comparison relationship between the two
returns in the evaluation process. Therefore, the key is to
take advantage of field awareness in the KDD process.
The incorporation of field awareness is subject
to how it can be signifying and filled in to the knowledge
discovery process. Ontology-based field awareness
representation, transformation, and mapping between
business and data mining systems is one of the proper
approaches to form field awareness.
c) Collaboration Among Human beings and Mining
Systems
The genuine requirements for discovering actionable knowledge in restraint-based framework is
more expected to be human involved rather than
automated. Human involvement is embodied through
the collaboration among humans (including users and
business analysts, essentially domain experts) and data
mining systems. This is accomplished through the
complementation
between
human
qualitative
brainpower, such as field awareness and field
supervision, and mining quantitative brainpower like
computational potential. Therefore, real-world data
mining possibly presents as a human-machinecooperated interactive knowledge discovery process.
For example, skills, metaknowledge, and
invented philosophy of field experts can lead or help out
with the selection of characteristics and models, adding
industry features into the modelling, generating highquality
assumptions,
designing
interestingness
measures by injecting business concerns, and rapidly
estimating mining results. This help basically can
progress the effectiveness and competence of drawing
out actionable knowledge.
d) Mining Exhaustively Patterns
Sometime there is a situation that numerous
mined patterns are attention-grabbing more to data
miners than to businesspersons and such type of
situations slowed down the deployment and
implementation of data mining in real applications. For
that reason, it is vital to estimate the actionability of a
pattern in addition to to further find out actionable
patterns to hold up smarter and more effectual decision
making. This leads to exhaustively pattern mining.
Mining exhaustively patterns should think as
how to get better both scientific and business
interestingness in the previous restraint-based
framework. Technically, it could be through enhancing
or generating more effective interestingness measures.
Additional awareness has to be remunerated to
business desires, intentions, field acquaintance, and
qualitative intelligence of field experts for their impact on
taking out deep patterns.
e) Improving Knowledge Actionability
Patterns that are motivating to data miners may
not guide essentially to business reimbursement, if
deployed. For instance, a large number of association
rules often are found, while most of them might not be
workable in business state of affairs. These rules are
generic patterns or technically interesting rules. Further
actionability upgrading is mandatory for producing
actionable patterns which is practically useful to
commerce.
The measurement of actionable patterns is to
follow the actionablilty of a pattern. Both technical and
business interestingness measures must be satisfied
from both objective and subjective point of view. For
those generic patterns identified based on technical
measures, business interestingness needs to be
checked and emphasized so that the business
requirements and user preference can be put into
proper consideration.
f)
Loop - clogged repetitive Improvement
Actionable knowledge discovery in a restraint based framework is probably to be a clogged rather
© 2011 Global Journals Inc. (US)
67
Global Journal of Computer Science and Technology Volume XI Issue XXIII Version I
through multiform restraints in addition to domain
knowledge.
December 2011
Data Driven Data Mining to Domain Driven Data Mining
December 2011
Data Driven Data Mining to Domain Driven Data Mining
than an open course of action. It includes repetitive
feedback to varying phases such as sampling,
assumption, feature selection, modelling, evaluation,
and interpretation in a human-involved approach. On
the other hand, real-world mining process is highly
repetitive, because the evaluation and refinement of
features, models, and outcomes cannot be completed
once but, rather, is based on repetitive feedback and
interaction before reaching the concluding juncture of
knowledge and decision-support report delivery.
The previous key elements indicate that real-world
data mining cannot be dealt by means of just an
algorithm; rather, it is truly essential to assemble a
suitable data mining infrastructure in order to find out
actionable knowledge from restraint-based situations in
a loop-clogged repetitive manner.
Global Journal of Computer Science and Technology Volume XI Issue XXIII Version I
68 g) Interactional and Concurrent Mining Supports
To support domain-driven data mining, it is
noteworthy to develop interactional mining supports for
human-mining dealings and to estimate the findings.
And also concurrent mining supports often are
necessary and can deeply promote the real-world data
mining performance.
For interactional mining supports, clever agents
and service-oriented computing are a number of highquality technologies. They can support flexible,
business-friendly, and user-oriented human-mining
interaction through building facilities for user modeling;
user knowledge achievement; domain knowledge
modeling;
personalized
user
services
and
recommendation; run-time supports; and mediation and
management of user roles, interaction, security, and
cooperation.
The facilities for interactional and concurrent
mining supports largely can improve the performance of
real-world data mining in aspects such as humanmining interaction and cooperation, user modeling,
domain knowledge capturing, reducing computation
complexity, and so forth. They are a few crucial
ingredients of subsequent generation of KDD
infrastructure.
VII.
CONCLUSION
Real-world data mining applications have
projected critical desires for discovering actionable
knowledge especially for real-users and industry needs.
Actionable knowledge discovery is significant and also
very challenging. It is listed as one of great challenges of
KDD. The research on this issue has latent to
revolutionize the alive state of affairs in which a huge
quantity of rules are mined but still few of them are
interesting to business, and to endorse the extensive
deployment of data mining into business.
This research paper had tried to provide a novel
data mining methodology referred to as Domain-Driven
Data Mining. It provides a systematic indication of the
issues in discovering actionable knowledge and
© 2011 Global Journals Inc. (US)
advocates the methodology of mining actionable
knowledge in restraint-based framework through human
mining system cooperation in a loop-clogged repetitive
refinement manner. It is useful for promoting the
paradigm shift from data-driven hidden pattern mining
to domain-driven actionable data discovery.
REFERENCES REFERENCES REFERENCIAS
1. Boulicaut, J-F., & Jeudy, B. (2005). Constraintbased data mining. In O. Maimon, & L. Rokach
(Eds.), The data mining and knowledge discovery
handbook (pp. 399–416). New York: Springer.
2. Omiecinski, E. (2003). Alternative interest measures
for mining associations. IEEE Transactions on
Knowledge and Data Engineering, 15, 57–69.
3. Pohle, C. (n.d.). Integrating and updating domain
knowledge with data mining. Retrieved from
http://citeseer.ist.psu.edu/668556.html
4. S. Sharma and K. Osei-Bryson, Role of Human
Intelligence in Domain Driven Data Mining In: Data
Mining for Business Applications, New York:
Springer Science+Business Media, 2009.
5. Proceedings of the 15th ACM SIGKDD international
conference on Knowledge discovery and data
mining 2009, Paris, France, June 28 - July 01, 2009.
6. J. Han and M. Kamber, Data Mining: Concepts and
Techniques, 2nd edition, London: Morgan
Kaufmann, 2006.
7. H. Varian, Intermediate Microeconomics Fourth
Edition,New York: W. W. Norton & Company, 1996.